Complete News World

Google is developing an AI that is able to compose music from descriptive text

The Mountain View giant intends to revolutionize the music industry and share the spotlight conquered by ChatGPT-3, although for now it has no intention of giving access to the general public until some “ethical” disputes are resolved.

The artificial intelligence (AI) prominence of ChatGPT-3, a chatbot signed by OpenAI, is already known. However, Google already has a new AI capable of picking up some of its success. Giant Mountain View has published a new article about an artificial intelligence model capable of producing high-quality music from any type of short text description. Google’s new music sensation was dubbed MusicLM; And although its capabilities have surprised company officials, they don’t plan to make it available to the general public until some issues are resolved.

High fidelity music

This music creation tool consists, according to business sources, of SoundStream, W2v-BERT, and MuLan, which are organized in what they call a hierarchical sequential training task. In this regard, the results obtained are remarkably good. In fact, the company has commented that a little explanation is enough to get a good result. Just ask MusicLM for a “smooth violin melody accompanied by a distorted guitar riff” and MusicLM will begin creating music “at 24kHz that remains flat for several minutes”.

He could also be asked to turn Pablo Picasso’s Guernica pain into a melody, “a mixture of reggaeton and EDM, with another spatial sound”, or something consonant with 1950s disco. . but, this is not every thing.

MusicLM can “transform whistling and humming melodies according to the method described in a text commentary”. At this time, the form cannot create recognizable sounds to accompany these rhythms and melodies, but it is much more advanced than similar tools capable of creating multi-minute compositions with an internal coherence.

See also  China sets up research development lab for its space program

Ethical issues to be resolved

To achieve these results, the AI ​​underwent extensive training with an archive of 2,80,000 hours of music to learn how to create coherent songs to describe “great complexity”. However, for now, the aspiration to revolutionize the music industry with this tool will remain on standby.

The document posted by Google states that “we have no plans to release templates at this time.” The company argued that there was a risk of creative content plagiarism, cultural appropriation, or misrepresentation. It seems that until these ethical issues are resolved, Google will not take a step forward.